Nvidia "Turing" GTX 11xx-1080 Megathread OFF TOPIC Edition

Discussion in 'Video Cards & Monitors' started by nope, May 19, 2018.

  1. nope

    nope Member

    Joined:
    May 8, 2012
    Messages:
    1,448
    Location:
    nsw meth coast


    [​IMG]
    https://wccftech.com/nvidia-geforce-next-gen-graphics-card-pcb-leak/

    One of NVIDIA’s Major OEM Partners Allegedly Returned 300,000 GPUs, Major Inventory Issues Being Cited For Next-Gen GeForce Launch Delay


    https://wccftech.com/nvidia-oem-partner-300k-gpu-inventory-issues-next-gen-geforce-delay/

    NVIDIA seems to be having lots of trouble managing their GPU inventory or the excess of it as highlighted in new reports from Taiwanese based sources. Now the inventory issues may have exploded at NVIDIA HQ since a major Taiwanese OEM has allegedly returned 300,000 GPU units back to NVIDIA, that may end up causing a delay for the next-gen GeForce parts.

    Could NVIDIA GPU Inventory Issues Delay Next-Gen GeForce Launch? Major Taiwanese OEM Allegedly Returned 300,000 GPUs
    In a report published by SeekingAlpha, it is stated the one major OEM who also falls in the “Taiwan Top 3” has returned 300,000 GPUs back to NVIDIA. Following is the summary of the report which is also quoting other sources:

    [​IMG]RELATEDNVIDIA’s Next-Gen GeForce Rumored To Feature New GPU Clock Algorithm, HDMI 2.1, RTX Tech and More – Product Briefing Allegedly Given To NVIDIA Board Partners

    Reports out of Taiwan now suggest that Nvidia has a gaming GPU inventory problem.

    ‘Semiaccurate’ reported on the issue yesterday, and cited excess inventory in the channel as the primary reason for new gaming GPU delay.

    The glut is so severe that one top Asian OEM partner reportedly returned 300k GPUs to Nvidia.

    This idea was discussed in more depth with members of my private investing community, The Razor’s Edge.

    via Seeking Alpha

    The issue is not that there’s no supply available for OEMs, the issue is that they have actually manufactured an inventory which they cannot manage now. That’s why one of their biggest original equipment manufacturing partner had to return a mass total of 300,000 GPUs using NVIDIA’s current GPU architecture, aka Pascal.

    [​IMG]

    Now there could be several reasons why NVIDIA produced such a large quantity of GPUs which is causing the inventory issues right now. First of all, Jensen wanted to return back to gaming market as soon as the cryptocurrency market started slowing down. Following is Jensen’s statement from February 2018:

    [​IMG]RELATEDNVIDIA Officially Launches GeForce GTX 1050 With 3 GB VRAM – Improved Core Specifications Along With A Cut Down Memory Interface

    We’re working really hard to get GPU down to the marketplace for the gamers and we’re doing everything to advise retailers and system builders to serve the gamers. And so, we’re doing everything we can, but I think the most important thing is we just got to catching for supply.

    Jen-Hsun Huang – NVIDIA’s CEO

    Gamers had been deprived of GeForce 10 gaming cards for several quarters after the mining spree started in 2017. The mining craze affected both NVIDIA and AMD, causing severe price inflations on gaming cards, and to top that up, the cards were in such a short supply that those who wanted to pay the hefty price for gaming cards had to wait for them to be actually available in retail & online outlets. We are talking about prices that reached over $2000 US for the Radeon RX Vega 64 and the GeForce GTX 1080 which retailed for $699 at launch.


    [​IMG]

    NVIDIA had their own estimates on when the mining craze would start dwindling and prices would return to normal. They had estimated that GPU prices would fall and start normalizing after Q3 ’18 however, we know now that prices are back at a reasonable level and it’s just Q2 ’18.

    “While NVIDIA was here they also let us know that the pricing in the market will continue to go up through Q3 of this year most likely before we start seeing any type of relief,” Hutch says.

    Unfortunately, the end to this is not right around the corner and we have not seen the worst of it yet,” writes Hutchins. He also went on to pledge that Massdrop will continue to list graphics card deals when it can and that it’s currently selling cards “right above our cost just barely covering back-end expenses.” Massdrop.com via PCGamer

    Strong demand in the cryptocurrency market exceeded our expectations. Cryptocurrency mining accounted for a higher percentage of revenue than the previous quarter. Keep in mind that’s very difficult for us to quantify down to the end customers view. Our main focus remains on our core market, as cryptocurrency will likely remain volatile.

    Colette Kress – NVIDIA’s Executive Vice President and Chief Financial Officer

    NVIDIA didn’t expect the price plunge so early or the mining craze to end this soon. Both AMD and NVIDIA had expected a 40% plunge in crypto demand by April, so they still kept manufacturing the GPUs, hoping for a new spree of mining to get users buying more cards. And so NVIDIA just kept on increasing supply.

    We don’t set prices at the end of the market, and the best way for us to solve this problem is, work and supply. The demand is great and it’s very likely the demand will remain great as we look through this quarter.

    [​IMG]

    And so, we just have to keep working on increasingly supply. We have our suppliers who are the worlds best and largest semiconductor manufacturers in the world and they’re responding correctly and I’m really grateful for everything they are doing, we got to do a catch up to that demand, which as such is really great.

    Jen-Hsun Huang – NVIDIA’s CEO

    The overestimation of the crypto demand and increased supply seemed to have backfired at NVIDIA. DigiTimes, in their latest report, highlights the dim prospects of crypto mining, so much so that Taiwanese suppliers of graphics cards are expecting major shipments and profit downfall in the current quarter.

    The sources said that the cryptocurrency mining heyday seen between April 2017 and March 2018 suffered an abrupt downturn in April due partly to the Bitcoin value plunging to under US$7,000 from a peak of nearly US$20,000 recorded in December 2017 and partly to governments of China, South Korea, the US and many European countries rushing to clamp down on digital-coin exchanges following exposures of scams, frauds and market manipulations.

    Taiwan graphic cards makers including Asustek Computer, Gigabyte Technology, Micro-Star International (MSI) and TUL have seen their inventories pick up significantly amid the drastic shrinkage in demand from cryptocurrency mining sector. But they have only slightly cut sales prices, maintaining gross margins at around 20%, which, though lower than the previous high of 40-50%, is still twice the level of 8-10% seen in early 2017.

    Accordingly, the sources expected the graphic cards supply chain to return in the second half of 2018 to the previous state of serving mainly the gaming sector seen before the rise of the cryptocurrency mining craze. And those makers with weaker deployments in the gaming sector may see their overall shipments of graphic cards fall significantly in the future.

    via DiGiTimes

    Due to crypto fall, NVIDIA OEM partners are returning the excess of supply. Now NVIDIA is known to have a tight grip on their top partners with programs such as NVIDIA Green Light and GPP (the highly controversial GeForce Partner Program which they pulled the plug from last month), but it looks like this time they have no option than to accept the GPU supply return. You might know that NVIDIA partners offer several GeForce SKUs such as the GTX 1060 with 5 GB memory and also a 3 GB variant of the GTX 1050. All of these are to reduce the huge inventory that NVIDIA and their partners currently have a hold of.

    [​IMG]
    NVIDIA enjoyed a successful Q1 FY19, beating estimates. OEM & IP got a major jump in revenue since it accounts for mining graphics cards too.

    It is also stated that NVIDIA is buying large supplies of GDDR5 memory to feature on graphics boards featuring their overstocked GPUs. NVIDIA might try to sell the cards themselves to limit inventory but 300,000 is a huge number and gamers have been waiting for two years since the previous GeForce gaming cards to get something new for their PCs. But this overstock can actually result in a short delay for the upcoming generation since who’ll buy older cards when new ones are just around the horizon?

    This would also explain why Jensen Huang, NVIDIA’s CEO, said during Computex 2018 that next-gen GeForce cards are a long way from now.

    [​IMG]

    Jensen has also previously insisted people to buy the GeForce GTX 1080 Ti that is the fastest gaming graphics cards to date but I personally have been using the card for over a year now and would love to upgrade my own PC. We have slight tidbits about the new GeForce cards but I would like NVIDIA to make some official statement sooner rather than later.

    Considering the alleged delay of next-gen cards, what would you do?


    https://wccftech.com/nvidia-next-generation-gpu-q3-18-power-logic/

    I saw a very interesting report published over at DigiTimes which cited an interview of Power Logic (a top tier cooling fan manufacturer that produces fans for the likes of EVGA) chairman Hsu Wen-Fang who stated that his company expects cooling fans sales to go through the roof starting Q3’18 because of next-generation GPUs launching around that time frame.

    Power Logic expects strong GPU cooler sales in Q3’18 from the launch of Next-Gen NVIDIA GPUs

    Power Logic is a known manufacturer of cooling fans of all types and sizes and even though no AIBs were named in the interview, we know that at least one major US AIB is their customer (you can see the chairman holding up an EVGA GPU in the file photo from DigiTimes below). This means that they would be privy to supply chain timelines and the certainty with which he expects Power Logic’s sales to go up translates to a buy order from one or all of these AIBs during that time frame.

    [​IMG]
    Power Logic chairman Hsu Wen-Feng holding up an EVGA GPU. Photo: Aaron Lee, Digitimes, June 2018

    NVIDIA has silently been preparing the launch of their brand new lineup for quite some time now and this is the second somewhat concrete source that we have heard tell about the expected time frame, the first being the Hot Chips presentation. Since Hot Chips presentations usually happen after the launch, the tentative timeline for launch now appears to be somewhere before that date. Before we go any further, here is the relevant extract from the interview:

    Cooling fan maker Power Logic expects to enjoy strong revenue growth momentum in the second half of 2018, mainly bolstered by stable orders for the firm’s new car ventilation fans and increasing GPU cooling fan shipments starting in the third quarter, according to company chairman Hsu Wen-feng.

    Power Logic’s major revenues are still generated by cooling fans for gaming PCs, which contribute up to 85% of the company’s annual revenues, with globally leading PC brand vendors all on its customer list, according to Hsu.

    – Hsu Wen-Feng interview to Digitimes, 2018.

    Keep in mind, however, that NVIDIA usually favors launch dates at the beginning/end of a quarter (so they can get the full financial impact in the next quarter) and that could translate to a launch happening sometime in July – which would also fit nicely with the Hot Chips presentation in August. Unless the company decides to screw convention and simply release architectural details at Hot Chips without actually launching the next-gen GPU lineup yet.


    The fact that Power Logic will start shipping coolers en masse in Q3’18 unfortunately does not help us pinpoint the exact month of launch, Considering we do not know how integrated the supply chains are with their American counterparts and depending on what kind of inventory mechanism they use (for eg, whether coolers are shipped months in advance or just-in-time). However, it does mean that NVIDIA will be launching its brand new series of graphics cards in a few months’ time.

    We still do not know for sure whether they will be calling it Turing/Ampere (or something else) but the company’s recent tweets related to the Turing institute do point towards the former. As far as nomenclature goes – that is anybody’s guess. The brand new GTX 1180/2080 graphics card (or whatever NV decides to call it) should offer a double-digit performance increment over the flagship of the last generation, the GTX 1080 Ti. Unfortunately, we have also heard rumors about the company experimenting with higher MSRP and price points to preserve the financial high they saw during the cryptocurrency boom but these remain mostly unsubstantiated at this point.


    :tongue:
    Launch soon
    The briefings of the board partners for the new cards were already running or running. The BoM is through, so now everyone can calculate for themselves (the time window I had already explained in my news) [​IMG]
    [​IMG]
    Unfortunately, this time I have been with geNDA't, so I can not rüberwursteln exact information this time. But my buddy is already scratching his feet, although he does not know anything yet. But as I know Wccftech so, he does not need much longer[​IMG]
    Update from 12.06.2018 18:40 clock

    In the meantime, we have unofficially learned from some board partners that Nvidia has already started training the relevant employees from the development departments. And since Nvidia does not prioritize any sales training courses for sales, the time calculation based on the tables listed below is likely to take on much more concise forms. If you follow the 3-month rule, the first board partner cards should appear on the market in late August or early September. However, some of the partners are now expecting a shift of at least two weeks, so that September seems rather plausible.

    Even if none of the partners are allowed to express themselves officially and may, the picture is thus condensed. We also have more information on technical details, but as long as we do not get it confirmed from at least one other source, we'll just say nothing about it. However, should this information prove true, AMD's graphics card business will certainly face tough times. The only lifeline is likely to be a fairer street price, because Nvidia will certainly not be the price trigger.

    Original News:

    What we officially know, of course, could be summed up quite quickly in a short sentence, because Nvidia itself understandably still covers badly. That's indulged, because in the end it's your product and its launch. But there are always one or two little puzzle particles, which in the sum then already gives a certain image, with which one can at least mentally play something. Back then, with Turing instead of amps, we were already in the right place and we were able to exclude a launch for the GDC and GTC in the spring due to our own research very early.

    [​IMG]

    Whether the name scheme now on a GeForce GTX 1180, 2080 or whatever may end up, is actually secondary, because it does not really much difference, as the child is now called. Much more interesting is the time of the rumored launch of reference cards and the board partner models. And that's exactly where we want to speculate something.

    [​IMG]

    Some board partners have meanwhile talked about their own model launches in August and September respectively. Add to this the (of course not officially confirmed) information that the bundling of the GPU and memory could start on June 15, 2018 and take a look at the following schedule, which roughly takes the time needed to produce a graphics card from the Material list, which outlines the development steps to the start of mass production and first delivery, then it could quite well fit.

    BOM release Bill Of Materials Release begin
    EVT Engineering Validation Test 1-2 weeks
    DVT Design Validation Test 2 weeks
    WS Working sample 1-2 weeks
    EMI Test Electromagnetic Interference Test less than a week
    PVT Production Validation Test 2-3 weeks
    PVT sorting
    PPBIOS Final BIOS a few days
    Ramp & MP Mass production and shipping a few days
    As I said, this is the "normal", generally held timing, a board partner also graduated in AMD cards in this form. Only Nvidia would not be Nvidia, if you did not want to have a little more control over the process. But we also like to weave in these little internals at this point. Nvidia calls the whole internally since Maxwell "Green Light Program" (also like flippantly called "Nvidia's Rules") and secures on the one hand the quality of the board partner designs and on the other hand avoids indirectly also possible RMA disputes in case of damage.

    You can see how you want it: Guidance or wise foresight. Does anyone remember the MSI GTX 660 Power Edition and the voltage cheat we uncovered ? Precisely for such reasons, Nvidia now incorporates further steps in the actual technological process, which, depending on the development progress, could then also cost one day or the other. Especially if details were objected. And that does not happen that often.

    Phase / Step
    description
    Specs and Guidelines
    Partners get the program guidelines and specifications
    CDP creation
    Partner submits CDP (Virtual Customer Design Project) to NVIDIA
    in accordance with the transmitted guidelines
    CDP approval NVIDIA reviews the CDP and approves it
    Design approval Partner submits mechanical design and NVidia reviews it
    sampling
    NVIDIA ships chip samples to partners.
    Acoustic approval
    NVIDIA, otherwise the partner provides the dB (A) curves
    Green Light approval
    The partner conducts the Green Light review using the special software (VBIOS and driver) and submits the result to Nvidia


    Option 1
    If the board passes Green Light, a Partner Production (PP) VBIOS will be provided by NVIDIA.

    Option 2
    If not, NVIDA wants to tell the partners what needs to be fixed. After fixing the issues this step starts again.

    Box art approval
    Partner provides the box art for Nvidia for review (using the preferred partner portal, must comply with the GeForce GTX brand guidelines)
    Mass production
    After all approvals the partner can start the mass production. No further changes of the Green Light approved design are allowed
    Why did we go so far? Suppliers also need a certain lead time, so that certain things can not be done on a case by case basis. For example, Asus and Gigabyte have the same PCB maker and Gigabyte and MSI have the same cooler ODM. To avoid collisions there, you can not put the time window too narrow. PNY as the responsible producer of reference cards (analogous to PC Partner for AMD) should already have all the necessary information. At least that means the current rumor situation.

    Our up-to-date information, which of course still causes a certain amount of blurriness due to certain delays in time, allows for the following speculations at the end :




      • Shipping of Turing bundles for FE from mid-June (Leak)
      • Launch of the Turing reference cards FE ("Founders Edition") in July (Leak, rumor)
      • Due to the generally slightly higher GDDR6 prices, the cards could be up to 100 euros more expensive in the EIA (assumption)
      • Launch of the Turing board partner cards in August / September (board partner statements, calculation example)
      • Launch of the first new Quadro Turing cards for Siggraph in August (suspected)
      • Performance higher with similar power consumption compared to the direct predecessor model (condensed rumor)
    Of course, these calculations are just pure mind games, but in every rumor and jigsaw piece, there's a bit of truth in the end. And even if we have not really become smarter as to the name scheme, then at least we know a little more about the production of graphics cards. And these are real, verifiable facts.

    https://www.tomshw.de/2018/06/12/nv...der-was-wir-ueber-turing-nicht-wissen-update/
    https://www.forum-3dcenter.org/vbulletin/showpost.php?p=11719150&postcount=1349




    https://videocardz.com/newz/jensen-huang-next-geforce-is-a-long-time-from-now



    During the first press conference before Computex, NVIDIA’s CEO Jensen Huang was asked multiple times when to expect the next GeForce series.

    It’s a long time from now
    What the conference has confirmed is that NVIDIA is not even ready to talk about the new series yet. Any question about the next installment in GeForce portfolio has been met with a laugh and no definite answer from Jensen.

    The press conference was an invite-only event. Those who were invited were welcomed by snacks handed personally out by Jensen. Unfortunately, those who were expecting a taste of a new Geforce were left with disappointment.

    No details about GeForce 20 were given, although journalists were trying to at least get a hint.

    What Jensen told everyone was that they will be invited to the launch of new GeForce, but they should not expect it very soon. This confirms what we have been hearing from the industry for the past few months, NVIDIA has not given any details to their partners yet, so there is nothing to talk about yet.
    EVERYTHING BELOW THIS LINE IS BULLSHIT

    _______________________________________________

    As usual take this with a grain of salt especially the "relative" performance

    1080 info:
    https://wccftech.com/nvidia-gtx-1180-custom-cards-by-august-september/
    Well we got a bit of a wait boys and grills
    https://wccftech.com/nvidia-geforce-gtx-1180-specs-performance-price-release-date-prelimenary/
    [​IMG]
    1180.jpg

    1070 info:
    [​IMG]
    1170.jpg
     
    Last edited: Jun 27, 2018
  2. Sledge

    Sledge Member

    Joined:
    Aug 22, 2002
    Messages:
    6,947
    Location:
    Adelaide
  3. dzajroo

    dzajroo Member

    Joined:
    Dec 23, 2010
    Messages:
    659
    guess we need to get the debate started with some controversial statements....

    I see what Nvidia did there.....when they launched 1080 founders edition for $100 more. It was all planing ahead. Now, they launch 1180/2080 for $699USD and say: "oh, but it's the same price as the old gen", problem is AIB partners will sell them for $699+ :D

    or they will just blame everything on mining/RAM/Inflation prices :D

    Regardless, the new GTX 1180/2080 (equality for everyone) will prob. only be on par or slightly faster than 1080Ti, then another half-year to a year wait for 1180/2080 Ti version.
     
  4. havabeer

    havabeer Member

    Joined:
    Dec 12, 2010
    Messages:
    3,601
    Or just buy the Titan xxx which normally comes out a little bit after release!
     
  5. Sledge

    Sledge Member

    Joined:
    Aug 22, 2002
    Messages:
    6,947
    Location:
    Adelaide
    At 3x the price....
     
  6. havabeer

    havabeer Member

    Joined:
    Dec 12, 2010
    Messages:
    3,601
    yeah thats the down side... i bought titan xp's when they where first released. i don't think i'll be doing that agian, will just wait till the 1180ti's come out
     
  7. OP
    OP
    nope

    nope Member

    Joined:
    May 8, 2012
    Messages:
    1,448
    Location:
    nsw meth coast
    See this shit is why i dont buy nvidia

    I had a 390x 390/290 before this for ages and a 770 and i was very happy with them

    IF they shit all over my 1080 performance wise (not optimizing drivers for cards a gen or so old as soon as the new ones come out) i am just gonna wait till 2020 and MCM 7nm+ amd stuff comes out. Nvidia are overcharging (mainly due to ram cartel bullshit but also due to AMD rtg dropping the ball)
    I bet we wont see those June cards till July at the earliest and August-Oct for AIBs
     
  8. dcl

    dcl Member

    Joined:
    Jan 26, 2002
    Messages:
    4,265
    Location:
    melbourne.au
    Does anyone know if the enthusiast gaming cards 1180/1170/etc will still be great for training neural networks with TensorFlow/Pytorch etc, or are nvidia going to strip that functionality out to force people to buy the tesla or quadro cards?
     
  9. HobartTas

    HobartTas Member

    Joined:
    Jun 22, 2006
    Messages:
    662
    Depends what you need exactly because if its FP64 then the consumer cards tend to have very little of it and I still have a couple of original Titans and the only consumer cards with higher scores than those are the 7990 and the Titan V/Black, in addition consumer cards are also lacking in not having ECC memory while tesla or quadro cards do have it.
     
  10. kot0005

    kot0005 Member

    Joined:
    Nov 11, 2011
    Messages:
    364
    Location:
    melbourne
    Nvidia taking their sweet time because AMD has nothing....almost 2 years since 1080 release..
     
  11. dcl

    dcl Member

    Joined:
    Jan 26, 2002
    Messages:
    4,265
    Location:
    melbourne.au
    FP64 and ECC memory isn't required. I was more worried they'd just disable some functionality with firmware/drivers.
     
  12. SnooP-WiggleS

    SnooP-WiggleS Member

    Joined:
    Sep 20, 2004
    Messages:
    2,867
    Location:
    Ballarat, Victoria
    I hope they do the second which is the reality. At a guess they might go 12gb instead of 8, and GDDR6 is 25% reported to be more expensive than GDDR5 (which is also inflated along with all other DRAMs etc). So a $100 premium doesn't suprise me, Nvidia knows their loyal fans will pay it anyway and gives them room to drop the price later when the market improves reducing the bad press of above MSRP prices. This video is good on memory prices in general.


    Here's hoping this card is more than just a 1080 ti die shrink and handles vulkan/dx12 better than pascal. So bit faster than a 1080 ti for dx11, but also seeing gains like the titan V does for Doom or Wolfenstein New Colossis (vulkan) over 1080 ti.

    They normally take about 2 years between generations (Maxwell 2014, Pascel 2016), so a launch in a couple of months is only a couple of months behind historically. Also they've been waiting on GDDR6 to happen as well.
     
  13. HobartTas

    HobartTas Member

    Joined:
    Jun 22, 2006
    Messages:
    662
    Don't think this will happen for two reasons and the first is that typically all they need to do is restrict whatever they want in the hardware silicon and secondly they are also pushing video cards as compute cards (CUDA) which basically that is what their current Titan is as its not really that crash hot as a gaming card and secondly if Nvidia cripples that too much there's always AMD who are ready to step in and sell their cards instead.
     
  14. OP
    OP
    nope

    nope Member

    Joined:
    May 8, 2012
    Messages:
    1,448
    Location:
    nsw meth coast

    never thought i'd see DF cite currytech but there you go
     
  15. hawpinghaxbag

    hawpinghaxbag Member

    Joined:
    Jun 6, 2005
    Messages:
    4,316
    Location:
    Your mum's place
    Depends how many vram controllers and the vram chip density. 256bit can be 4/8/16GB
     
  16. OP
    OP
    nope

    nope Member

    Joined:
    May 8, 2012
    Messages:
    1,448
    Location:
    nsw meth coast
    16gb will be for highend vr/4k+ cards
     
  17. Mathuisella

    Mathuisella Member

    Joined:
    Dec 3, 2002
    Messages:
    5,319
    Location:
    in your gearbox...grindin
    not sure why we aren't doing higher bus width than 256bit, i remember 10 years ago a 384bit wide memory bus. why not 512 or higher ?
     
    tunksy likes this.
  18. hawpinghaxbag

    hawpinghaxbag Member

    Joined:
    Jun 6, 2005
    Messages:
    4,316
    Location:
    Your mum's place
    Cost i believe, AMD did 512bit gddr but their cards are more bandwidth hungry which is why they hopped on the HBM trainwreak
    384bit or 352bit in the 1080ti's case is reserved for the higher end cards
     
  19. OP
    OP
    nope

    nope Member

    Joined:
    May 8, 2012
    Messages:
    1,448
    Location:
    nsw meth coast
    this, its all about raw ghz these days anyway with gddr6 tipped to hit 16gbps+
     
  20. Sledge

    Sledge Member

    Joined:
    Aug 22, 2002
    Messages:
    6,947
    Location:
    Adelaide

Share This Page